- Anticipate what’s next — and act with confidence
- Success stories
- Healthcare manufacturer transforms HR enquiry management with AI
Healthcare manufacturer transforms HR enquiry management with AI
HR productivity increased
Experts spend less time answering routine questions
Employee engagement boosted
HR Answers available instantly, 24/7
A strategic solution
Multilingual AI chatbot can be deployed across multiple functions and geographies
Challenge
With more than 26,000 employees in over 30 countries, this pharmaceutical and chemical manufacturer wanted to harness GenAI in an HR chatbot. By automating answers to employee questions, 24/7 support could be introduced to improve employee engagement, provide immediate answers to questions, and improve HR operational efficiency.
The company set itself a very high target for success. To encourage employees to use and trust the system, the new AI chatbot would need to:
- Drive dynamic conversations in real time, mimicking human interactions without seeming contrived
- Integrate seamlessly with text-based documents including Word and PDF files, ensuring complete understanding and processing of information
- Offer a user-friendly experience with intuitive navigation and answer questions in natural language
As an advanced GenAI application, the platform would be trained to answer a broad range of routine and unusual employee questions about pay, holiday, sickness benefits and many others. Answers would be derived from a broad range of raw text HR documents, but the solution would learn, adapt and evolve to keep up with latest requirements.
To avoid the need for periodic investment in technology, the company chose a SaaS deployment on cloud with GCP tooling throughout. GCP managed the cloud deployment, but the company also needed help to develop an engaging user experience UX and to manage the data that would drive the chatbot. With its deep understanding of GenAI and practical experience of building successful AI chatbots GFT was chosen to lead the project.
Engagement
GFT and GCP initially produced a proof of concept (PoC) to showcase the capabilities of Generative AI on the Google Cloud Platform. This comprised a question/answer retrieval system populated by over 70 HR procedure documents. The chatbot agent focussed on US HR policy-related inquiries and was rigorously tested by employees, who posed challenging questions and provided detailed feedback. Results we analysed and collated by GFT so the model could be adapted and fine-tuned.
The PoC demonstrated the immense potential of GenAI to improve employee engagement by answering questions quickly and accurately. The company asked GFT to develop a production-ready chatbot for use by its 10,000 US employees from January 2024. To encourage use, employees had to be able to ask questions in English.
The production chatbot included robust security guardrails to prevent the large language model “hallucinating” or answering inappropriate questions. The chatbot uses a RAG (Retrieval Augmented Generation) framework to ensure that results are based on the most recent, relevant data that can be trusted. RAG also reduces the need to train the model continuously on new data and update its parameters as circumstances evolve.
To simplify the user experience, user identification and sign-on was provided by Security Assertion Mark-up Language (SAML) federation in GCP. This ensures that only those who are entitled to use the systems can do so.
GFT maintained the service, monitored performance, and logged change requests. Following the successful PoC, a series of regular updates was scheduled to enhance the service and refine the user experience.
Benefit
The new chatbot was well received by users. Employees are spared of the need to peruse lengthy HR documents and procedures manuals. The chatbot answers both routine and unusual questions and users appreciate the speed and accuracy of answers. The chatbot also offers the company savings of around three FTEs, who can focus on tasks that add more value.
The success of the chatbot lies in its ability to integrate the immense power of GenAI in a simple chat interface. Users can choose to ask routine or unusual questions and a pre-trained Google’s Generative AI (LLM) provides answers based on relevant information provided by a previous open-source information retriever. The Retriever reduces the context provided to the LLM to improve the accuracy of answers.
One of the strategic benefits of the chosen methodology is that the LLM does not need to be trained with private datasets. It can simply be fed with raw text extracted from HR policies and other corporate documents, including Word and PDF files. This means that the chatbot can easily be redeployed in other function and geographies. GFT is currently collaborating with the client on a new Spanish language version of the chatbot.
Download - just one click away
-
Success story
Healthcare manufacturer transforms HR enquiry management with AI
